The historical definition of NAFLD, created in the 1980s, was widely used until 2022 [3] when Eslam turned the negative definition (i.e., nonalcoholic) into a positive diagnostic criterion: MAFLD [4]. This development was continued in 2023, when a novel definition was coined: MASLD, which requires concurrent steatosis and at least one cardio-metabolic criterion [5]. This effort also accounts for the patients with MASLD, who consume moderate amounts of alcohol. This condition is termed metabolic dysfunction-associated alcohol-related liver disease (MetALD) [6].

Several scholars have highlighted the limitations of the MASLD [7, 8] and MetALD definitions [9] while supporting head-to-head comparative studies [10]. When comparisons have been made, they often suggested that MAFLD may more selectively identify patient populations at risk of severe liver-related and extra-hepatic outcomes [11]. Conversely, MASLD is more inclusive and captures larger strata of patients [11].

MASLD affects 38% of adults and up to 14% of children and adolescents, and the prevalence rate in adults is projected to exceed 55% by 2040 [12]. Groups at high risk of MASLD comprise those with obesity, type 2 diabetes (T2D), and other features of the metabolic syndrome [13]. The fact that MASLD often develops secondary to specific endocrinopathies is increasingly being recognized [14].

MASLD is modified by gender and hormonal status. It is more common in men than in premenopausal women while postmenopausal women’s MASLD rates are similar to men’s [15]. On the other hand, women have a higher risk of advanced fibrosis than men, especially after the age of 50 years [16].

Currently, in the US, MASLD is the leading indication for liver transplantation in women and subjects with hepatocellular carcinoma (HCC) [12]. However, only a subset of subjects with MASLD will progress to advanced forms of liver disease, and the leading cause of mortality among MASLD patients is cardiovascular disease (CVD) [12]. Like other liver disease etiologies, surveillance for early detection of HCC is warranted among those with MASLD and advanced fibrosis, cirrhosis, or portal hypertension [17].

Further to liver-related outcomes and to CVD, accumulating data suggest that MASLD may be associated with certain extra-hepatic cancers, such as those of the gastrointestinal tract and of the urinary system [18, 19]. The stage of liver fibrosis and gender modulate the risk of hepatic and extra-hepatic outcomes [20, 21]. A growing body of studies supports chronic kidney disease as another extra-hepatic outcome whose risk parallels the severity of MASLD [22]. The presence of both conditions also increases the risk of CVD [23].

Although Mendelian-randomization studies do not invariably support a cause-and-effect relationship between MASLD and these extra-hepatic outcomes [24, 25], clinicians should be aware of the risks of CVD and extra-hepatic cancers among those with MASLD as a part of a holistic approach to this systemic disorder [13, 26].

Hepatocytes are the central hub of lipid metabolism of the human body since they take up and secrete lipids into the bloodstream, carry out de novo lipogenesis as well as degrade lipids via β-oxidation. An imbalance in these pathways results in intrahepatic accumulation of neutral fat that becomes evident as steatosis [27] and can be detected histologically as well as by multiple imaging methods [28]. While ultrasound is most widely used in the clinical routine, magnetic resonance imaging with assessment of proton density fat fraction is the most accurate non-invasive method [28].

Unlike the adipose tissue, the liver is not physiologically devoted to the accumulation of fat and the accumulation of intrahepatic fat has the potential for triggering lipotoxic phenomena such as cell degeneration (i.e., ballooning) and cell death. The ongoing stress may trigger pro-inflammatory and pro-fibrotic cascades and thereby lead to metabolic dysfunction-associated steatohepatitis (MASH), an advanced form of MASLD, and subsequently to MASH-fibrosis or MASH-cirrhosis [29]. MASLD results from an intricate organ crosstalk with adipose tissue and the intestinal tract playing key roles [30]. The former acts as a supplier of free fatty acids (FFAs) and of hormonal substances, such as leptin and adiponectin that affect the hepatocellular lipid handling [31]. The intestinal tract is tightly connected to the liver via portal circulation and both organs work closely in the digestion and handling of nutrients. However, MASLD patients often display an increased gut permeability and intestinal dysbiosis that leads to portal circulation of gut-derived toxins and hepatic inflammation [32].

Steatogenic and fibrotic progression of MASH are intimately and bi-directionally associated with insulin resistance (IR) in the context of systemic metabolic dysfunction in subjects with either features or the full-blown metabolic syndrome [33]. Additionally, age, sex, reproductive status, hormonal influxes, and lifestyle factors (diet, smoking, alcohol, physical exercise, sedentary behavior) are key modifiers of the distribution of body fat, of intrahepatic balance of pro- and anti-lipogenic, inflammatory, and profibrogenic pathways [34]. Notably, MASLD displays a strong heritable trait, and genetic studies shaped our understanding of MASLD pathogenesis [35]. On a population level, a variant in patatin-like phospholipase domain-containing 3 (PNPLA3) constitutes the strongest genetic contributor to MASLD and highlights the importance of lipid droplet handling within hepatocytes, which is supported by other MASLD-related variants such as MBOAT7 [35]. In particular, the PNPLA3 I148M seems to accumulate on lipid droplets and to impair their degradation [36]. However, genetic studies also highlight the complexity of the condition with processes like secretion of very low-density lipoproteins (TM6SF2, ERLIN1), lipogenesis (GCKR, ApoE), diversion of triglycerides (MTARC1), and many others being affected [35].

Bile acids (BAs), produced from cholesterol in hepatocytes, are involved not only in bile formation and fat digestion, but also serve as signaling molecules involved in metabolic processes such as lipid and glucose homeostasis [37, 38]. Disruption of BA homeostasis leads to the progression of MASLD and development of MASLD-HCC [39]. Recent studies have found that interleukin-1 receptor 1 (IL1R1), gut microbiota, and BA changes synergistically contribute to the development of MASLD [40], suggesting pathogenic cross-talks between BA metabolism, gut dysbiosis, and “metaflammation”, i.e., sterile metabolic inflammation, in MASH. BAs are elevated in MASLD, particularly ursodeoxycholic acid (UDCA), taurocholic acid (TCA), chenodeoxycholic acid (CDCA), taurochenodeoxycholic acid (TCDCA), and glycocholic acid (GCA). However, BA profiles vary owing to geographic regions or disease severities [41]. Notably, TCA, TCDCA, taurolithocholic acids, and glycolithocholic acids exhibit a potential ability to identify MASH [41]. Trafficking of cholesterol to mitochondria through steroidogenic acute regulatory protein 1 (STARD1), the rate-limiting step in the alternative pathway of BA generation, not only promotes the progression from steatosis to MASH by sensitizing hepatocytes to inflammatory cytokines-induced cell injury [42] but also plays a key role in the pathogenesis of MASLD-HCC (Figure 1) [37]. This is particularly relevant for MASH and HCC development as the synthesis of BA through the classic pathway controlled by CYP7A1 is repressed by endoplasmic reticulum (ER) stress. Finally, fibroblast growth factor (FGF) 21, a potent regulator of glucose and lipid homeostasis, acts as a negative regulator of BA synthesis [43], paving the way for targeted pharmacological interventions in MASLD/MASH arena, as discussed below. Another pathogenically and therapeutically relevant alteration in MASLD/MASH is dysfunctional farnesoid X receptor (FXR)-FGF19 feedback signaling. This leads to elevated BA production, higher pericentral biliary pressure, and pericentral micro-cholestasis that may account for raised GGT serum concentrations in a proportion of MASLD/MASH individuals [38, 44, 45].

Obesity, unhealthy diet and sedentary lifestyle are key contributors to both T2D and MASLD. In line with that, lifestyle interventions aiming at dietary changes and increased physical activity constitute an established therapeutic approach [46]. It has been repeatedly demonstrated that weight loss of 5–7% can reverse steatosis while higher weight loss can reverse fibrosis. However, many participants undergoing lifestyle intervention fail to lose weight, and many regain weight after the end of the study [46]. Because of that, many advocate for bariatric surgery as the more reliable method to achieve long-term weight loss [47]. While this invasive method is associated with potential side effects such as dysphagia, abdominal discomfort, or nutritional deficiencies, large studies with long-term follow-up demonstrated that it decreases the overall mortality as well as the rate of major adverse cardiovascular events (MACEs) and major adverse liver outcomes [48, 49]. As an alternative approach, several weight-reducing drugs have been recently developed. Among them, glucagon-like peptide 1 receptor agonists (GLP-1 RAs) and sodium-glucose co-transport 2 inhibitors (SGLT2is) are commonly used in individuals with diabetes and are associated with reduced MACEs and overall mortality [50–52]. Among GLP-1 RAs, semaglutide was tested in a phase 2 randomized, double-blind, placebo-controlled trial (RCT), showing a higher rate of resolution of MASH than placebo [53]. These results were strengthened by a phase 3 RCT where semaglutide improved liver histology in more patients than the placebo arm [54]. Similarly, encouraging results were reported for SGLT2is, in that dapagliflozin outperformed placebo in improvement of histological MASH in a recent RCT [55]. While GLP-1 RAs remain the best studied class of drugs leading to weight loss due to decreased food intake, their effect can be strengthened by combination with glucagon receptor and/or glucose-dependent insulinotropic polypeptide agonists. Among such drugs, the dual agonists tirzepatide and survolutide were able to resolve MASH without worsening fibrosis in the corresponding phase 2 RCTs [56, 57]. These data remarkably demonstrate the usefulness of weight-reducing approaches in MASLD. While these drugs show beneficial extrahepatic effects, none of these agents are currently approved for treatment of MASLD and their prescription is currently reimbursed only in patients with diabetes in most countries.

As the leading liver disease worldwide, MASLD became a coveted pharmaceutical target and yielded unprecedented insights into the relevance of several signaling cascades. Recently, resmetirom, a liver-directed, thyroid hormone receptor beta (THR-β)-selective agonist, became the first FDA-approved MASH agent and highlighted the importance of endocrine signaling in MASLD [58]. Earlier studies demonstrated impaired THR-β function and the corresponding decrease in β-oxidation of fatty acids in MAFLD. The approval of resmetirom is based on a phase 3 RCT where it was superior to placebo in both MASH resolution and improvement of liver fibrosis [59]. FGF21 is another hormone regulating glucose and lipid metabolism as well as insulin sensitivity. While FGF21 agonist pegozafermin led to improvement in liver fibrosis in a phase 2b RCT, the FGF21 agonist efruxifermin did not reduce fibrosis in subjects with MASH-related cirrhosis [60, 61]. Peroxisome proliferator-activated receptors (PPARs) belong to widely studied nuclear receptors that play an important role in liver metabolism, inflammation, and fibrogenesis. Their broad hepatoprotective effects are substantiated by recent clinical trials that led to approval of two PPAR agonists for treatment of primary biliary cholangitis (PBC) [62, 63]. In a phase 2b RCT, the pan-PPAR agonist lanifibranor significantly improved a histological MASH score without worsening liver fibrosis [64].

The advancement of liver-directed RNA silencing approaches (RNAi) [65] opened the possibility of personalized treatment of MAFLD directed at the key genetic risk factors. In that sense, the treatment with the PNPLA3 RNAi JNJ-75220795 resulted in decreased liver steatosis in two independent phase 1 RCTs [66], while the HSD17β13 RNAi associated with improved alanine aminotransferase levels [67]. Finally, a repurposing of FDA-approved drugs might also be beneficial in MASLD as shown in a recent RCT that reported a decreased steatosis in subjects treated with low-dose aspirin for 6 months [68].

Finally, since existing findings indicated that the type rather than the amount of fat emerged as an instigator of lipotoxicity and sensitization to advanced stages of MASLD, with cholesterol playing a key role [42, 69, 70], it has been shown that statins may be another pharmaceutical approach for MASH. In this regard, Yun et al. [71] reported that statins are significantly associated with reduced liver-related event risk in patients with MASLD, especially among those with elevated alanine aminotransferase levels, suggesting a viable preventive strategy for such a population.

Until recently there was a clear distinction between patients suffering from MASLD and patients with ALD, which was mainly based on the average daily or weekly alcohol intake (old definitions: NAFLD: women < 10 g of pure ethanol per day; men < 20 g of pure ethanol per day; ALD: women > 50 g of pure ethanol per day; men > 60 g of pure ethanol per day) [13]. However, this differentiation excluded a large group of (overweight) patients showing clear signs of steatotic liver disease, who were not consuming alcohol in amounts that would be considered ALD patients, but who were ingesting an amount of alcohol that would exceed the criteria for MASLD (women: 20–50 g, men: 30–60 g) [5]. Indeed, it is rather common for high-calorie foods to be consumed alongside moderate and even higher alcohol intake. Moreover, it has been suggested that surveys on self-reported alcohol consumption may be similar to energy reporting in overweight and obese individuals [72–74]. Together, this may result in misleading estimations of intake and subsequently of the prevalence of MetALD. Therefore, the recent estimates of the National Health and Nutrition Examination Survey (NHANES) 2017–2020 consider that steatotic liver disease was prevalent in 37.85% of individuals, of whom 32.45% were diagnosed with MASLD, 1.17% with ALD, and 2.56% with MetALD [75]. Also, data from the NHANES 2017–2018 have revealed marked disparities in the prevalence of MetALD and ALD in the US, which were related to ethnicity and living circumstances, e.g., food safety, suggesting the involvement of other factors [76, 77]. Using data from the UK Biobank, Schneider et al. [78] reported that 10.8% of SLD cases could be attributed to MetALD, indicating that the prevalence of MetALD may be higher in Europe than in the US. Interestingly, the average alcohol intake per person in individuals over 15 years has been reported to be rather similar between UK (9.7–10.7 L in 2024–2025) and the US (8.7–9.97 L in 2024–2025) [79, 80], while the prevalence of overweight or obesity seems to be even higher in the US than in the UK [81]. Stockwell et al. [82, 83] demonstrated in several studies that self-completed questionnaires using quantity-frequency and graduated-frequency methods may be subject to significant underreporting, which can be partially overcome by combining these assessments with an evaluation of beverage-specific yesterday consumption (BSY). However, it needs to be noted that the latter methods do not capture long-term drinking patterns. Taken together, further studies assessing the prevalence of MetALD among patients with MASLD are needed using other, more objective tools to complement 24-h recalls, food frequency questionnaires, and CAGE and AUDIT-C assessments of alcohol intake, in order to achieve a better overview of prevalence and related risk factors. Indeed, in recent years, biomarkers of alcohol intake such as phosphatidylethanol, ethyl glucuronide, ethyl sulfate, and fatty acid ethyl esters have become widely available and have been reported to be useful additions when assessing alcohol ingestion [84–86]; however, they, too, may only reflect alcohol intake in the recent past.

As the amount of alcohol consumed differs markedly between patients with ALD (alcohol intake women: > 350 g/week, men: > 420 g/week) and patients with MetALD (alcohol intake women: > 140 g/week, men: > 210 g/week), there is an ongoing discussion as to whether these patients require a different treatment, especially since patients with ALD may suffer from the condition without additional cardiometabolic risk factors while these are present in patients with MetALD [87]. Indeed, while MetALD and ALD are both chronic, progressive, and potentially terminal liver diseases, they are also markedly different. For instance, acute alcohol-associated steatohepatitis, which is found in heavy and long-term drinkers, is characterized by significant inflammation, deranged liver function, and high mortality rates (20–50% at 28 days in severe cases) [88, 89]. Data on the prevalence of acute steatohepatitis in MetALD patients and its outcome are scarce. Nevertheless, it has been demonstrated that in women, MetALD is associated with a higher hazard for all-cause mortality (+83%), which is thought to stem from their greater susceptibility to developing more severe ALD at lower levels of alcohol intake [90]. This has been linked to differences in gastric alcohol metabolism by alcohol dehydrogenase (ADH), lower body water, and estrogen-related differences in the susceptibility to intestinal barrier dysfunction and macrophage activation by lipopolysaccharide (LPS) [91–94]. Moreover, excessive alcohol intake in MASLD patients has been shown to be associated with a higher mortality risk than in non-drinkers [95]. Furthermore, especially extended binge drinking (at least 13 days/year) has been found to be related to a higher risk of mortality in a survey in the US [96]. However, as the number of studies on MetALD patients and models is still limited, no specific treatment besides abstaining from alcohol and changing lifestyle is currently available for the treatment of MetALD.

As acknowledged in many publications, MASLD and ALD are two clearly different pathological conditions (for an overview, see [97, 98]). It has also been suggested by the results of many studies that MASLD and ALD share several clinical and mechanistic features [97, 98]. For instance, disturbances in lipid metabolism resulting in intracellular lipid accumulation in hepatocytes have been shown to be integral in both MASLD and ALD [99]. Furthermore, in both diseases, this accumulation of lipids has been associated with ER stress and mitochondrial dysfunction, ultimately resulting in cell damage and death (Figure 1) [100]. The activation of hepatic stellate cells (HSCs) triggers not only further inflammatory alterations but also enhances collagen synthesis and deposition, which has been linked to hepatocellular damage in both diseases [101]. Additionally, alongside the observed alterations in the liver lipid synthesis and subsequent stellate cell activation, both MASLD and ALD have also been associated with changes in intestinal microbiota composition and intestinal barrier dysfunction, subsequently leading to increased translocation of bacterial toxins as well as viral compounds [102, 103]. The latter have been linked to an activation of Toll-like receptor (TLR)-dependent signaling cascades, which lead to inflammation but also promote lipid accumulation and stellate cell activation in liver tissue. Indeed, in model organisms, the treatment with antibiotics has been shown to diminish both the development of MASLD and ALD [104, 105]. Whether a combination of a lifestyle/genetic factor promoting the development of MASLD and ALD exacerbated the different alterations, e.g., the disturbance of lipid metabolism, ER stress, and intestinal barrier dysfunction through additive or synergistic effects remains to be determined. The results of recent studies suggest that in the setting of IR, which is frequently found in MASLD patients, alcohol metabolism via ADH-dependent metabolism is impaired [106, 107]. Indeed, in these studies, it was shown that through TNFα- and c-Jun N-terminal kinase (JNK)-dependent signaling cascades, the phosphorylation status of ADH in liver tissue is altered, subsequently leading to a loss of enzyme activity. It was further shown that in mice with MASLD, alcohol clearance was significantly lower than in MASLD-free control mice, further suggesting that alcohol is maintained longer in the system, which could evoke further damage [106]. This needs to be determined in larger human studies, as it may also have implications not only with respect to organ damage but also cognition and subsequently the use of machinery and behavior in traffic. Although gastric ADH may be an important player in alcohol metabolism and hence a determinant of the availability of alcohol in the circulation, once reaching the liver, the oxidative metabolism of alcohol is accounted for by cytochrome P450 2E1, whose expression is induced by alcohol consumption. Nevertheless, the implication of the impaired alcohol metabolism via ADH in MetALD in humans deserves further investigation.

In summary, the interaction of MASLD and alcohol consumption is still poorly understood, despite affecting a large yet unknown number of patients with steatotic liver diseases in many countries. Further studies are needed to assess the number of individuals affected using better, more reliable tools. Studies in patients and animals with MASLD have reported impaired ADH activity; however, the impact of this impairment on alcohol elimination in humans remains to be determined. Furthermore, there is limited data assessing the interaction between MASLD and elevated alcohol consumption at the organ and cellular levels. As our understanding of the molecular mechanisms is still limited, there are few therapies beyond lifestyle interventions such as abstinence and weight reduction.

BAD occurs when BA reabsorption in the ileum is defective or feedback regulation via the FXR-FGF19 axis is impaired. Primary BAD, often underdiagnosed, is linked to inadequate FGF19 signaling, leading to unchecked CYP7A1 activity and hepatic BA overproduction [154]. Secondary BAD results from ileal resection, Crohn’s disease, or cholecystectomy.

Diagnostics include serum 7α-hydroxy-4-cholesten-3-one (C4), with values > 48 ng/mL suggestive of BAD, and fasting FGF19 < 145 pg/mL. The SeHCAT retention test has excellent sensitivity and specificity. Fecal BA quantification remains a gold standard but is cumbersome [155].

First-line therapy is BAs sequestrants, though tolerability is limited. FXR agonists and FGF19 analogs represent emerging therapies that target the underlying pathophysiology [156].

Cholestasis results in the intrahepatic retention of hepatotoxic BAs. In PBC, UDCA is standard, but many patients are incomplete responders [157]. Obeticholic acid is approved as a second-line therapy but is limited by pruritus. In primary sclerosing cholangitis (PSC), norUDCA showed significant benefit in a 2025 phase 3 trial [158]. Recent findings indicated an increased expression of STARD1 in patients with PBC, suggesting that the mitochondrial alternative pathway of BA synthesis could play a role in PBC pathogenesis [159].

In pediatric cholestasis, ileal bile acid transporter (IBAT) inhibitors such as maralixibat and odevixibat are approved, reducing pruritus and serum BA levels [160].

IBD is characterized by altered BA pools with reduced secondary BAs, weakening FXR/TGR5 signaling. This contributes to barrier dysfunction and inflammation. Experimental studies show that BA supplementation or intestine-restricted FXR agonists restore epithelial barrier integrity and modulate immunity [149, 161]. Microbiome-targeted therapies are also under investigation [162].

Primary conjugated BAs such as taurocholate stimulate spore germination, while secondary BAs such as DCA and LCA inhibit growth. Antibiotic-induced dysbiosis reduces secondary BAs, predisposing to CDI. Fecal microbiota transplantation restores BA profiles and reduces recurrence [146, 162].

Secondary BAs, particularly DCA, promote tumorigenesis by inducing oxidative stress, DNA damage, and Wnt/β-catenin signaling [163]. Epidemiological studies show higher fecal BA load in Western diets and CRC patients [152]. Loss of FXR in colonic tissue amplifies tumorigenesis.

As mentioned above, in MASLD, BA metabolism is profoundly altered, with elevated conjugated and 12α-hydroxylated BAs. Dysregulated FXR-FGF19 signaling leads to excessive BA synthesis, while taurocholate activates stellate cells via S1PR2, promoting fibrosis [150, 164, 165]. Dysbiosis shifts microbial BA metabolism toward pro-fibrogenic profiles, worsening inflammation [151].

Serum conjugated BAs (GUDCA, GCDCA, TCDCA) correlate with histologic severity and fibrosis. Therapeutics including FXR agonists, FGF19 analogs, and resmetirom (THR-β agonist approved in 2024), illustrate how BA modulation is central to MASLD management [166–168].

Multiple serum, fecal, and imaging-based biomarkers have emerged to quantify BA metabolism. Serum C4 and FGF19 are useful for diagnosing BAD, while serum BA panels predict fibrosis in MASLD [150, 151, 155]. Fecal BA profiling characterizes IBD, CRC, and CDI, while SeHCAT is highly sensitive for BAD diagnosis [155, 165].

BA-targeted therapies range from UDCA to novel small molecules, biologics, and microbiome interventions. NorUDCA showed benefit in PSC, while obeticholic acid, cilofexor, and tropifexor reduce fibrosis but are limited by side effects [157, 158, 167]. FGF19 analogs and engineered mRNA restore BA signaling but raise safety concerns. IBAT inhibitors are approved in pediatric cholestasis [160]. Microbiome-directed therapies, including FMT, restore secondary BA production in CDI. Resmetirom, a THR-β agonist, was FDA-approved in 2024 for MASH [168], but whether this agent affects BA homeostasis or not remains to be established.

Challenges include assay standardization, balancing therapeutic efficacy with safety, and addressing patient heterogeneity [167–169]. Microbiome therapy outcomes remain variable, and long-term safety of FXR agonists and FGF19 analogs requires further study. Future directions include plasma BA panels in clinical practice, approval of norUDCA, development of microbiome-based BA modulators, and AI-driven BA modeling for personalized therapy.

BAs have emerged as central modulators of digestive and metabolic diseases. Advances in mechanistic understanding, biomarker discovery, and targeted therapies highlight their dual role as pathogenic drivers and therapeutic opportunities. In MASLD, specific BA signatures contribute to fibrogenesis, making them attractive diagnostic and therapeutic targets. The next decade promises integration of BA-focused strategies into precision hepatology and gastroenterology.

Global projections estimate that liver cancer incidence may rise from approximately 0.87 million in 2022 to over 1.5 million by 2050 if current prevention efforts stagnate [277]. Although the burden from chronic viral hepatitis is gradually decreasing due to vaccination and antiviral therapy, cases linked to MASLD, alcohol consumption, and aging are sharply increasing [12, 278]. The recent redefinition from NAFLD/NASH to MASLD/MASH more accurately reflects the metabolic and inflammatory basis of disease and has reframed HCC risk models, highlighting that a significant fraction of HCC cases can develop even in the absence of cirrhosis—complicating surveillance paradigms [279].

In contrast, the incidence of intrahepatic CCA (iCCA) has shown a consistent increase in Western countries over the past two decades, while perihilar and distal CCA [extrahepatic CCA (eCCA)] rates have remained stable or declined slightly [280]. Recent population-based analyses from England confirm a continuing rise in iCCA incidence and mortality, largely driven by metabolic comorbidities, aging, and improved diagnostic imaging [281]. Major risk factors for CCA include PSC, hepatolithiasis, recurrent cholangitis, biliary parasitic infections (Opisthorchis viverrini, Clonorchis sinensis), and fibrotic or metabolic liver disease. Molecularly, FGFR2 fusions and IDH1/2 mutations characterize iCCA, whereas HER2 amplification/overexpression and KRAS/BRAF alterations predominate in eCCA [282]. The interplay of environmental and genomic factors underscores CCA’s heterogeneity and its growing overlap with the metabolic and viral risk spectrum of HCC.

Collectively, these shifting patterns underscore a global epidemiologic transition in adult liver cancers, where viral hepatitis-driven carcinogenesis is gradually being replaced by metabolic and aging-related mechanisms. This evolution not only alters the demographic and risk landscape but also calls for revised surveillance strategies and biomarker-driven prevention models tailored to non-viral populations.

Hepatoblastoma (HB) is the most common primary liver malignancy of children, with the majority (90%) of cases diagnosed at under 5 years of age. Although HB is a rare cancer, with an incidence of 1.7 cases per million children, cases have risen in the past few decades, possibly as a consequence of improved survival of premature and low-weight births [283]. HB arises from a small number of genetic modifications of which β-catenin activating mutations are found in the majority of cases, and alterations in the 11p15.5 locus occurring in around 85% of tumours. However, HB is a phenotypically heterogeneous cancer with tumours evolving across a continuum of cellular states arising from a combination of clonal evolution and epigenetic plasticity [284].

Updated EASL (2024) and AASLD (2023–2025) guidelines now emphasize risk-stratified surveillance and treatment allocation [285, 286]. The BCLC 2022 strategy continues to dominate staging, integrating refinements in intermediate-stage management and multidisciplinary decision-making. The LI-RADS v2024 update introduced key improvements in ultrasound-based surveillance, CT/MRI response assessment, and standardized interpretation post-locoregional therapies such as yttrium-90 (Y-90) and stereotactic body radiotherapy (SBRT) [287]. Despite strong recommendations, surveillance underuse remains widespread, particularly in MASLD populations where ultrasound sensitivity is poor. This has driven interest in biomarker-augmented surveillance models, such as the GALAD panel and cell-free DNA methylation assays, alongside risk stratification tools like aMAP and PAGE-B, which are being validated for individualized surveillance frequency and modality [288–290].

The EASL-ILCA Clinical Practice Guidelines (2023) for iCCA have introduced molecularly informed recommendations for diagnosis, staging, and therapeutic allocation, emphasizing systematic next-generation sequencing (NGS) for actionable targets [IDH1, FGFR2, BRAF, HER2, microsatellite instability (MSI)-H/TMB-high, NTRK] and multidisciplinary tumor boards [291]. The 2025 EASL guidelines on perihilar and distal CCA refine imaging, surgical margins, and systemic therapy management, distinguishing iCCA from eCCA both biologically and clinically [292]. The ESMO 2023–2024 consensus similarly advocates for upfront molecular profiling to inform first-line and subsequent treatment decisions [282, 293]. Collectively, these updates reflect a shift from purely anatomic to biology-driven stratification, paralleling the precision-medicine paradigm already transforming HCC.

In advanced HCC, atezolizumab-bevacizumab (IMbrave150) and tremelimumab-durvalumab (STRIDE) remain standard first-line regimens [277, 294–296]. The HIMALAYA long-term results confirmed durable survival benefit and manageable safety over more than four years, particularly in patients at risk for bleeding. In contrast, lenvatinib-pembrolizumab (LEAP-002) and cabozantinib-atezolizumab (COSMIC-312) failed to improve overall survival, underscoring the need for biologically informed combinations rather than empirical TKI-IO pairings [297–300]. In 2024, the EMERALD-1 trial established the first positive immunoembolization strategy: Durvalumab plus bevacizumab with TACE significantly improved progression-free survival versus TACE alone, heralding a new era of locoregional-immunotherapy synergy [301]. In the adjuvant space, IMbrave050 demonstrated improved recurrence-free survival with atezolizumab-bevacizumab after curative therapy [302].

For CCA, immunotherapy has become a new standard. The TOPAZ-1 phase III trial showed that durvalumab + gemcitabine/cisplatin improved overall survival compared with chemotherapy alone (median OS 12.8 vs. 11.5 months; HR 0.80), establishing the first IO-chemo combination in biliary tract cancers [303]. Similarly, KEYNOTE-966 confirmed OS benefit with pembrolizumab + gem-cis (HR 0.83; OS 12.7 vs. 10.9 months) [304]. These results cement immune-chemotherapy backbones as the new front-line standard. In MSI-H or TMB-high CCA, anti-PD-(L)1 monotherapy remains recommended per tissue-agnostic approvals [282].

At the locoregional level, selective use of TARE/TACE or HAIC for unresectable iCCA is expanding, with translational studies exploring immune activation and vascular remodeling as combinatorial endpoints, mirroring progress seen in HCC.

Single-cell and spatial multi-omics analyses have redefined the HCC microenvironment, exposing an intricate tumor-immune-stromal ecosystem [305, 306]. Malignant hepatocytes reprogram tumor-associated macrophages (TAMs) into immunosuppressive TREM2-positive phenotypes, which restrict CD8⁺ T cell infiltration. In preclinical models, TREM2 blockade reactivates antitumor immunity and synergizes with PD-(L)1 inhibition, offering a plausible mechanism behind the clinical heterogeneity in immunotherapy response [307, 308]. Preclinical MASLD-HCC models reveal dysfunctional, metabolically exhausted T cells and myeloid reprogramming that impede immunosurveillance, aligning with clinical observations of attenuated immune checkpoint inhibition (ICI) efficacy in MASLD-related HCC [309, 310]. A deeper understanding of the features of the HCC tumour microenvironment that predict sensitivity to atezolizumab-bevacizumab can enable precision approaches for the selection of ICI. A recent scRNAseq investigation revealed three infiltrating cell types associated with around 40% of responders; these comprised two types of CD8⁺ T cell (CD8 Temra and CD8 Tex) and pro-inflammatory CXCL10⁺ macrophages [311]. The other 60% of responders characteristically carried features of genomic instability associated with anti-VEGF response indicative of an “Angiogenesis-driven” subset of patients. Disease progression was mainly associated with low immune infiltrations with enrichment of Notch signalling and progenitor-like states. However, a subset of resistant tumours was identified as immune-enriched but infiltrated by immunosuppressive CD14⁺ monocytes and TREM-2⁺ macrophages.

In iCCA, single-cell and spatial transcriptomics have mapped distinct immune niches within a dense desmoplastic stroma populated by cancer-associated fibroblasts (CAFs) driving immunosuppression via TGF-β, CXCL12, and IL-6 signaling [312, 313]. CAF-tumor crosstalk activates PI3K-AKT/Notch pathways, supporting epithelial-mesenchymal plasticity and therapeutic resistance [314]. Murine models highlight the contribution of cancer stem-like cells and myeloid-epithelial interactions to treatment escape and metastasis [315]. Together, these findings underscore the stroma-dominated immune exclusion of CCA and the potential for “stroma-first” strategies that parallel emerging approaches in HCC.

Mechanistic studies since 2022 have identified several tumor-intrinsic vulnerabilities. The ferroptosis axis (GPX4-SLC7A11) represents a key therapeutic target; ferroptosis inducers combined with TKIs or ICIs potentiate tumor death in preclinical models, offering a strategy to overcome sorafenib resistance [316]. Concurrently, Hippo-YAP/TAZ-TEAD signaling and focal adhesion kinase (FAK) pathways are being targeted to modulate tumor stiffness, stemness, and immune exclusion. Inhibition of FAK improves anti-PD-1 efficacy and reshapes the fibrotic stroma in mouse models, laying the groundwork for ‘stroma-first’ clinical approaches [317].

CCA shares similar vulnerabilities. Intrahepatic models demonstrate that ferroptosis resistance—via overexpression of GPX4, SLC7A11, and ACSL3—is associated with poor prognosis and metabolic reprogramming [318]. YAP/TAZ activation sustains proliferation, stemness, and immune evasion in iCCA [319], and its crosstalk with FAK-mechanosignaling contributes to stromal stiffness and tumor invasion [320]. Targeting these pathways may overcome the desmoplastic barrier that limits immune infiltration and drug delivery—suggesting strong parallels between HCC and CCA in mechanobiology-driven immunoresistance.

Together, these data highlight mechanotransduction and ferroptosis resistance as shared therapeutic vulnerabilities across HCC and CCA. Translationally, targeting these pathways could serve as a unifying strategy to overcome immune exclusion and stromal stiffness, paving the way for “stroma-focused” precision therapies in liver cancer.

The only curative treatment for primary liver cancers is surgery, specifically either tumour resection or a liver transplant, the latter arguably being optimal since, in addition to removing detectable tumours, it also removes any unseen micrometastatic lesions and the diseased liver in which the cancer developed. However, up to 70% of cancers treated by surgery recur in both HCC and CAA, including extra-hepatic as well as intra-hepatic growths [321, 322]. This high rate of recurrence is due to a combination of contributory factors including circulating tumour cells, residual intrahepatic tumour cells, or micrometastases, the pro-oncogenic environment of the diseased liver and an immunosuppressed tissue microenvironment, which can persist due to intra-hepatic but also extra-hepatic influences such as the dysregulated gut microbiome [323]. The role of immunity in recurrence is supported by studies that report high circulating platelet-to-lymphocyte ratio (PLR) and/or neutrophil-to-lymphocyte ratio (NLR) with early recurrence in HCC [324]. Elevated PLR is also unfavourable after surgery for CCA [325]. Hence, immunotherapy is being investigated for perioperative approaches with the aim of enhancing anti-tumour immunity for the prevention or delay of tumour recurrence. The IMbrave050 trial examined adjuvant atezolizumab-bevacizumab in the setting of high-risk recurrence HCC patients undergoing tumour resection. While initial results were reported as encouraging, with a 28% reduction of recurrence [302], an updated report of the trial failed to demonstrate a durable benefit [326]. Ongoing trials of adjuvant anti-PD-1 agents (EMERALD-2 and CheckMate-9DX) will help further determine if adjuvant ICI treatment has merit. As CCA is an immunogenic tumour [327], there is good justification for perioperative immune checkpoint therapy in those patients who are eligible for surgery. The ACCORD trial combined post-surgical chemoradiation with camrelizumab (anti-PD-1) and demonstrated OS and RFS efficacy [328].

Neoadjuvant protocols are also being tested in ongoing trials and may have greater efficacy due to the presence of a fully intact tumour antigen-immune axis. The aims of neoadjuvant treatment include pre-operative tumour downsizing, bridging to transplant, and prevention/delay of recurrence. In HCC, MORPHEUS-NEO HCC is evaluating immunotherapy combinations (atezolizumab, bevacizumab, tiragolumab, tobemstomig) as neoadjuvants in resectable HCC [NCT05908786] while PRIME-HCC is investigating the safety and efficacy of nivolumab/ipilimumab combination treatment prior to tumour resection [329]. Similar studies in resectable CCA are determining neoadjuvant treatments with durvalumab plus tremelimumab (anti-CTLA-4) combined with gemcitabine [NCT04989218], tsilezumab (anti-PD-1) plus gemcitabine and cisplatin [NCT06903273], and adebrelimab (anti-PD-L1) plus lenvatanib [NCT06208462].

Genomic and immunologic subclassification is now guiding patient stratification. CTNNB1 (β-catenin)-mutant HCC defines an immune-excluded phenotype associated with poor ICI response; combination strategies (STING agonists, YAP/FAK modulators) are under preclinical and early-phase clinical evaluation [330]. The cGAS-STING pathway, in turn, mediates antigen spreading and myeloid activation when coupled with radiation or oncolytic agents, suggesting translational synergy between RT-STING-IO triads [331].

Molecularly targeted therapy has reshaped iCCA management. Futibatinib (FOENIX-CCA2) achieved durable responses in FGFR2 fusion-positive iCCA, leading to global regulatory approval [332]. Pemigatinib (FIGHT-202) and derazantinib maintain efficacy across updated cohorts [333]. Ivosidenib improved PFS and OS in IDH1-mutant iCCA (ClarIDHy) with sustained long-term benefit [334]. HER2-directed therapies—trastuzumab deruxtecan and zanidatamab—achieve clinically meaningful responses in HER2-amplified CCA [335]. For rare NTRK fusions, larotrectinib and entrectinib remain tissue-agnostic options [336].

Emerging clinical trials are testing IO + targeted therapy combinations (FGFR/IDH inhibitors plus anti-PD-1 or anti-VEGF), while biomarker-guided enrichment strategies are integrating transcriptomic and immune profiling to tailor patient selection. Together, these efforts signify the translational shift toward precision oncology in CCA, comparable to HCC’s biology-informed trial design.

Genomic studies confirm hepatitis B virus (HBV) DNA integration and TERT-promoter mutations as early carcinogenic events and potential biomarkers for minimal residual disease surveillance [337, 338]. Circulating HBV-host junctional DNA and ctDNA methylation assays are now entering prospective trials for post-resection monitoring. Parallel research on the gut microbiome-liver axis shows distinct microbial signatures correlating with ICI efficacy, prompting ongoing efforts to integrate microbiome modulation into immunotherapy regimens [339].

Liquid biopsy approaches in CCA are advancing across multiple analytes: plasma ctDNA, bile cfDNA, and extracellular vesicles enhance diagnostic sensitivity in indeterminate biliary strictures [340, 341]. Panels of exosomal circRNAs show promise for early diagnosis and recurrence monitoring in pilot cohorts [342]. Standardization of preanalytical workflows and prospective multicenter validation remain urgent priorities for clinical translation.

Advanced patient-derived organoids (PDOs), 3D spheroids, and microfluidic liver-on-chip systems now recapitulate tumor heterogeneity, fibrotic microenvironments, and immune cell interactions, enabling more predictive preclinical screening [343, 344]. In vivo, MASLD-faithful murine models are clarifying immune-metabolic crosstalk and testing ferroptosis, stroma-targeted, and myeloid reprogramming interventions. Translationally, GPC3-directed CAR-T and bispecific T-cell engagers, as well as personalized neoantigen DNA vaccines, have shown promising preclinical and early clinical signals, supporting the expansion of immunotherapy beyond checkpoint blockade [345, 346].

iCCA and eCCA organoids, co-culture epithelial-stromal models, and biliary organ-on-chip platforms replicate desmoplasia and plasticity, enabling testing of FGFR/IDH + IO and stroma-targeted combinations with immune and mechanical readouts [312]. Integration of single-cell and transcriptomic data into preclinical pipelines is paving the way for transcriptomic subtype-guided clinical trials.

Artificial intelligence (AI) has rapidly emerged as a transformative tool in HCC research, integrating imaging, genomics, and clinical data to enhance diagnostic and prognostic precision. A recent scientific metric analysis encompassing nearly 4,000 publications reported a 12-fold increase in AI-related liver cancer studies between 2013 and 2022, with predominant applications in imaging-based diagnosis (37%), predictive modeling (19%), and molecular bioinformatics (18%) [347]. Deep learning-based radiomics now enable automated detection and characterization of focal liver lesions across CT, MRI, and ultrasound modalities, achieving diagnostic accuracies exceeding 90% in differentiating malignant from benign lesions [348]. In a large-scale study including over 26,000 ultrasonographic images, convolutional neural networks reached a specificity of 97% for malignancy detection [349]. Beyond diagnostic applications, AI-driven models for survival and recurrence prediction have shown promising performance, with ensemble machine learning algorithms attaining accuracies up to 92% for outcome prediction across different HCC stages [350]. Nonetheless, systematic reviews consistently highlight the limited prospective and multicentric validation of these models, which currently constrains their clinical implementation [351]. As large annotated datasets and federated learning frameworks continue to expand, AI is expected to complement conventional and molecular risk stratification approaches, fostering precision surveillance and personalized therapeutic decision-making in liver cancer.

AI applications are also expanding rapidly in CCA, a heterogeneous and often late-diagnosed malignancy of the biliary tract. Over the past decade, research on AI in CCA has grown significantly, particularly in the fields of diagnostic imaging, preoperative risk assessment, and recurrence prediction [352, 353]. Radiomics and deep learning-based algorithms have demonstrated strong capabilities in distinguishing iCCA from HCC and benign hepatic lesions, with area-under-the-curve values frequently exceeding 0.90 in retrospective studies [354]. AI models trained on ultrasonographic and cross-sectional imaging datasets have also shown high diagnostic accuracy, achieving sensitivities of up to 92% in multicenter analyses [349]. Beyond imaging, integrative AI classifiers that combine transcriptomic signatures with clinical parameters are being developed to predict lymph node metastasis and postoperative recurrence risk; however, most remain at early stages of validation and lack large-scale, multicentric testing. Importantly, clinical translation is beginning to take shape, as illustrated by ongoing trials evaluating AI-guided biopsy approaches in suspected CCA (NCT05374122) [https://clinicaltrials.gov/study/NCT05374122]. Despite these encouraging advances, significant challenges persist—including limited access to standardized, annotated datasets, inter-institutional variability in imaging protocols, and the “black-box” nature of deep learning models. Future research should prioritize multi-omics data integration, development of explainable AI frameworks, and prospective validation studies to advance precision diagnosis and individualized therapy in CCA.

Beyond its diagnostic and prognostic roles, AI holds promise as a translational integrator, capable of merging imaging-derived features with genomic and immune signatures. This convergence could enable real-time, biology-informed clinical decision-making, further bridging the gap between bench discoveries and patient care in liver cancer.

The field of liver cancer is undergoing a convergence of clinical innovation and mechanistic insight. Modern therapy is shifting from empiricism to biology-guided precision—anchored in detailed molecular and immune profiling. The integration of spatial and single-cell data, ferroptosis and stromal mechanobiology, and immune-etiologic diversity is informing both rational combination design and risk-based clinical trial enrichment. Future success will depend on embedding translational endpoints into trials, standardizing MASLD-oriented preclinical models, and bridging molecular discoveries with pragmatic clinical benefit.

CRC is the third most common malignancy and the second leading cause of cancer-related death worldwide. In particular, patients with metastatic stage IV CRC have a very poor prognosis [355]. CRC has served as a paradigm for multistep carcinogenesis with accumulating mutations in APC, KRAS, BRAF, p53, as well as components of the TGF-β and PI3K signaling pathways such as SMAD4 and PIK3CA. Germline mutations in APC are the cause of the familial adenomatous polyposis (FAP) cancer syndrome, an autosomal dominant inherited condition with the formation of numerous colon polyps early in life. In patients with Lynch syndrome (also called hereditary nonpolyposis CRC or HNPCC), which is the most common inherited cancer syndrome in humans, mutations in genes of the DNA mismatch repair system (in particular Mlh1 and Msh2) have been identified [355]. Furthermore, CRC served as a paradigm for the definition of the Immunoscore, a tumor classification system that is based on the composition of immune cells in the stroma and their mode of infiltration [356, 357]. Here, we summarize important current advances and challenges in CRC research and therapy.

Advances in CRC research and therapy also entail new challenges. PDOs, for example, are a valuable tool for pre-screening of treatment responses, but standardized culture protocols are still lacking. They also incompletely model the complex immune microenvironment of CRC. Patient-derived xenografts, preferably in host mice with humanized immune systems, would be a better option, but due to cost and infrastructure constraints, it is unrealistic for them to become a routine clinical tool. Another unsolved challenge is gender-specific medicine for CRC. Gender differences in CRC development, treatment, and treatment-related toxicity have long been recognized. Possible explanations in metabolism, immune responses, hormonal status, and the importance of certain cancer genes on sex chromosomes have been elucidated, but have not yet led to gender-specific, tailored therapies. Furthermore, major challenges remain in the treatment of CRC and there is a worrying increase in CRC incidence rates in people under 50 years of age.

AI systems in endoscopy, imaging, pathology, and risk stratification are increasingly embedded in clinical workflows. Their benefits depend on large, often cross-border datasets; compliance with GDPR/HIPAA therefore requires privacy-by-design governance, clear purpose limitation, and auditable access control. Ethical review should appraise dataset provenance, re-identification risks, model drift monitoring, and subgroup performance reporting to prevent inequities in underrepresented populations [432–435].

Multiple meta-analyses confirm AI-assisted colonoscopy improves adenoma detection; however, a 2025 multicenter observational study suggested a potential “deskilling” effect—endoscopists exposed to routine AI performed worse when subsequently operating without AI. Editorials have urged competency preservation (periodic non-AI sessions, simulation, behavioral training) and post-deployment surveillance as adoption scales [429, 436, 437].

Omics-guided care in HCC, CRC, and MASLD/MASH enhances stratification and targeted therapy, but complicates consent (data reuse/storage, cross-border sharing, familial implications). CIOMS and Helsinki provide guardrails for risk-benefit balance, vulnerable groups, data/samples governance, biobanks, and post-trial access; protocols should use layered, literacy-appropriate consent and clarify policies for incidental/secondary findings [432, 433].

The ethical imperative is to avoid a two-tier system in which advanced diagnostics and therapies cluster in affluent centers. Equity-oriented reimbursement and procurement (tiered pricing, pooled purchasing), tele-expertise, and technology transfer can extend benefits to safety-net settings and low- and middle-income countries (LMICs); equity metrics should be tracked alongside clinical outcomes. A salient, recent example is Egypt’s HCV elimination program [438], which paired nationwide screening with locally manufactured or discounted direct-acting antivirals (DAA) and scaled treatment capacity. Independent evaluations document millions treated at sustainable cost and rapid reductions in HCV prevalence and sequelae—demonstrating that price negotiations, domestic production, and delivery at scale can operationalize affordability and equity.

Developers and institutions should document model intent, training data domains, and failure modes; clinicians should disclose AI involvement and maintain human oversight with clear override pathways. These measures support trust and align with responsible-innovation principles. A recent multicentre observational study in colonoscopy [436] reported a “deskilling” signal among endoscopists routinely using AI, with performance decrement when AI was withdrawn. The finding underscores the need for pre-specified monitoring plans, routine competency maintenance, incident reporting, and subgroup performance audits—practical accountability mechanisms that complement efficacy trials and help prevent unintended harms during post-deployment use.

Contemporary syntheses quantify the shifting aetiologic mix—declining viral hepatitis where vaccination/antivirals scale, and rising metabolic liver disease in parallel with obesity, diabetes, and alcohol—informing prevention, surveillance, and workforce planning [431]. MASLD guidance now recommends pragmatic, stepwise NIT-based case-finding in individuals with T2D/obesity, abnormal liver enzymes, or steatosis on imaging [13].

Translational epidemiology moves beyond accuracy to ask whether innovations improve hard outcomes at acceptable cost, across diverse care levels. For NITs, population-based and health-system studies evaluate referral yield, fibrosis progression prediction, and equity of access; for AI in endoscopy, program metrics (reach, adoption, fidelity, subgroup effects) complement sensitivity/specificity and adenoma detection rate (ADR) [429, 430].

Large cohorts and meta-analyses show that achieving sustained virological response (SVR) after DAA therapy reduces—but does not abolish—HCC risk, especially in advanced fibrosis/cirrhosis, justifying continued surveillance in high-risk subgroups [439–441]. For HBV, accumulating evidence suggests lower HCC risk with tenofovir vs entecavir in some populations, informing first-line antiviral choices and cost-effectiveness models [442–444].

Recurrent findings across digestive diseases—worse outcomes with lower income, education, rural residence, or minoritized status—argue for embedding equity targets into innovation programs (subsidized diagnostics, mobile elastography, community CRC screening with quality assurance). Epidemiologic monitoring should flag “innovation gaps” early (e.g., attenuated benefit of AI in bandwidth-constrained settings) to trigger corrective action [445–447].

Disability-adjusted life years (DALYs), mortality, and preventable fractions support prioritizing high-yield interventions—viral hepatitis elimination, fibrosis case-finding, and quality-assured CRC screening—before niche applications, with periodic reprioritization as epidemiology evolves [13, 431].

Policies should ensure NITs, essential antivirals, and proven screening AI reach safety-net and rural clinics (pooled procurement, tiered pricing, tele-mentoring). Implementation should include equity dashboards (utilization by socioeconomic stratum, geography, race/ethnicity) and remedial actions when gaps emerge [445–447].

Balance early detection against overdiagnosis anxiety and downstream harms; publish transparent thresholds (e.g., fibrosis/steatosis cut-offs), ensure pathways for confirmatory testing, and use shared decision-making aids adapted for health literacy.

Maintain human skills via credentialing, simulation, and deliberate “non-AI” exposure; deploy drift detection, human overrides, and incident reporting; disclose AI use to patients and to institutional review boards during evaluations [436, 437].

Respect local norms and legal frameworks (GDPR/HIPAA); use privacy-preserving technologies where feasible (de-identification with expert determination, data minimization, federated learning). CIOMS/Helsinki emphasizes public accountability, community engagement, and protections for vulnerable groups, which are vital for multi-center registries and biobanks [432–434].

In summary, innovation is accelerating across hepatology and gastroenterology—from AI-assisted detection and image-based elastography to omics-guided therapeutics. Epidemiology quantifies where need is greatest, validates real-world performance, and reveals inequities; ethics provides the guardrails—privacy, consent, fairness, transparency, and just access—so advances reduce rather than entrench disparities. A durable roadmap couples (1) global burden monitoring and equity-aware priority-setting; (2) stepwise, guideline-based deployment of validated NITs and AI with post-deployment safety/competency surveillance; (3) privacy-preserving, GDPR/HIPAA-compliant data governance; and (4) adherence to Helsinki/CIOMS principles throughout research and implementation. Done together, these pillars convert technological promise into measurable gains in survival, quality of life, and health equity for patients with liver and digestive diseases worldwide.